Researchers from the Massachusetts Institute of Technology (MIT) have developed an AI model for identifying ambiguous results in medical images.

According to the developers, even when using neural networks for diagnosing diseases from X-rays or MRI data, there is always a risk of errors. This is because images can be blurry or contain artifacts.

“Having a digital assistant can help in decision-making. The very fact of detecting uncertainty in an image can influence the doctor’s diagnosis,” noted MIT computer science candidate Marianna Rakich.

The method is based on segmentation – a process in which medical images are divided into separate areas and carefully examined to identify potentially dangerous areas.

Named after the Greek goddess of chance, the AI model Tyche uses Bayesian neural networks, which are able to handle uncertainty. These networks are trained on a set of medical images labeled by expert doctors.

The MIT neural network has several advantages compared to other AI methods:

• more accurately diagnoses diseases, as it takes into account uncertainty in the images;
• generates fewer false positives;
• helps doctors better understand how AI arrived at its conclusion.

The new model can have many applications in medicine. It can help doctors more accurately diagnose cancer at an early stage and predict the outcome of the disease, enabling them to make more informed decisions. Tyche also allows researchers to develop new treatment options.

Earlier, Google Cloud and German medical company Bayer announced the creation of an AI platform to help radiologists diagnose faster.

Recall that in March, scientists from the University of Ottawa announced the implementation of a neural network in their work to detect cardiovascular diseases.

Found a mistake in the text? Highlight it and press CTRL+ENTER